EEG-Based Classification of Motor Imagery Tasks Using Fractal Dimension and Neural Network for Brain-Computer Interface

• Published in: IEICE Transactions on Information and Systems Vol. E91.D; no. 1; pp. 44 - 53
• Main Authors: NAKAGAWA, Masahiro, PHOTHISONOTHAI, Montri
• Format: Journal Article
• Language: English
• Published: Oxford The Institute of Electronics, Information and Communication Engineers 01.01.2008; Oxford University Press
• Subjects: Applied sciences; Artificial intelligence; Biological and medical sciences; brain-computer interface (BCI); Computer science; control theory; systems; Computer systems and distributed systems. User interface; Computerized, statistical medical data processing and models in biomedicine; Connectionism. Neural networks; Electric, optical and optoelectronic circuits; electroencephalogram (EEG); Electronics; Exact sciences and technology; fractal dimension (FD); independent component analysis; Medical management aid. Diagnosis aid; Medical sciences; motor imagery; neural network; Neural networks; Software; Discriminant analysis; Automatic classification; fractal dimension (FD); brain-computer interface (BCI); Independent component analysis; Multiple layer; Man machine dialogue; motor imagery; Electroencephalography; Autoregressive model; Neural network; Signal classification; electroencephalogram (EEG); Backpropagation algorithm; Feedforward neural nets; Fractal dimension; User interface; Imaging; Comparative study; Fixed point
• ISSN: 0916-8532; 1745-1361
• DOI: 10.1093/ietisy/e91-d.1.44

In this study, we propose a method of classifying a spontaneous electroencephalogram (EEG) approach to a brain-computer interface. Ten subjects, aged 21-32 years, volunteered to imagine left-and right- hand movements. An independent component analysis based on a fixed-point algorithm is used to eliminate the activities found in the EEG signals. We use a fractal dimension value to reveal the embedded potential responses in the human brain. The different fractal dimension values between the relaxing and imaging periods are computed. Featured data is classified by a three-layer feed-forward neural network based on a simple backpropagation algorithm. Two conventional methods, namely, the use of the autoregressive (AR) model and the band power estimation (BPE) as features, and the linear discriminant analysis (LDA) as a classifier, are selected for comparison in this study. Experimental results show that the proposed method is more effective than the conventional methods.